Abstract: An offshore wind power prediction model combined with LSTM-FCN network is proposed, in which wind shear physical quantities are introduced into the data to more accurately predict offshore wind power generation. Two sets of wind turbine data from an offshore wind farm data website in Zenodo were selected for analysis and prediction verification. After standardized preprocessing of the dataset, AMLSTM-FCN network and CNN network, LSTM network, LSTM-FCN network were used to do the comparison experiments, in which AMLSTM-FCN network was predicted in 2 wind turbine data, RMSE, MAPE, MAE are respectively for No. 5 wind turbine: 6.9434, 14.01%, 48.6636, for No. 6 wind turbine: 2.6933, 7.12%, 17.2536, the data training network without wind shear data is used in the same time period. The obtained prediction results show that the prediction accuracy decreases from the four indexes. Experiments show that AMLSTM-FCN networks have higher prediction accuracy in offshore wind power prediction, and wind shear also has a significant impact on offshore wind power.